| Hi,
I like to follow up on this one, which I posted a while ago, since the cause of the issue has finally been identified.
It turns out that it was caused by uninitialised Ruta annotation variables. Ruta doesn’t seem to allow an annotation variable to be set null at the point of declaration or later, so we assumed that it would implicitly be initialised to null during script execution for every new CAS, but this is not the case in Ruta 2.6.1. The variable would hold a reference to an annotation from the previous CAS when it hasn’t been assigned a new value from the new CAS, because the rule that makes the assignment is never fired.
I have illustrated this using the attached example code where a CasRuntimeException is the result with Ruta 2.6.1, but the problem seems solved in Ruta 2.8.0. This exception could in our full setup sometimes trigger a NullpointerException during exception handling, which is what I first reported. I assume the problem would be gone in Ruta 3.0.0 as well.
I think the matter can be considered closed, although we are not through with all tests yet. I wonder though if it would generally be useful to allow the null value to be assigned to annotation variables.
|
Example_TS.xml
Description: XML document
RutaAnnotationVariableError.java
Description: Binary data
Hi Peter,
Ruta version is 2.7.0 with UIMA version 2.10.2 in this setup.
We are to best of my knowledge not trying to access any internal Ruta objects and there is nothing that makes any covered text assignments.
Hi, and which version of UIMA Ruta? Do you access internal Ruta objects somehow? Do you (by accident) try to assign covered text? Best, Peter Am 11.11.2019 um 10:23 schrieb Richard Eckart de Castilho: Hi Mario,
which version of the UIMA Java SDK are you using?
-- Richard
On 11. Nov 2019, at 09:58, Mario Juric <[email protected]> wrote:
Hi Peter,
A while ago we started to get some erratic null pointer exceptions from Ruta because the type of some feature structure element is null (see stack trace below). The error is not consistently reproducible, in fact it seldomly occurs and when reprocessing the document it doesn’t happen again. We therefore think there are some race conditions at play when running in a multithreaded environment as we do in production, and I was hoping that maybe you would get an idea what might be causing it just by looking at the stack trace.
Cheers
Mario
-- Dr. Peter Klügl R&D Text Mining/Machine Learning Averbis GmbH Salzstr. 15 79098 Freiburg Germany Fon: +49 761 708 394 0 Fax: +49 761 708 394 10 Email: [email protected]Web: https://averbis.comHeadquarters: Freiburg im Breisgau Register Court: Amtsgericht Freiburg im Breisgau, HRB 701080 Managing Directors: Dr. med. Philipp Daumke, Dr. Kornél Markó
|
